Any future affect recognition system would be presented with an enormous amount of data with broad scope. Our initial approach to developing experiments for affective data collection was to create one relatively straightforward experiment, and to control possible extraneous variables as much as possible. It was decided to focus on a "frustration" task; an experimental paradigm which induced negative arousal in the subject, and roughly corresponded to the kind of hardware or software problems that users encounter in everyday interaction with computers.
In our frustration experiment, we brought subjects in under the pretense that their task was to participate in a vision-oriented computer game. The game consisted of a series of puzzles, and the task was to click the mouse on the correct answer at the bottom of the screen to advance the screen to the next puzzle. Subjects received $10.00 for their participation, but the game was also a competition; the individual who received the best overall score and speed at the end of the data collection was awarded a $100.00 prize. Thus, an incentive was created to increase subjects' desire to play quickly and receive a good score.
An example of a typical puzzle in the frustration experiment.
However, what the subjects were not aware of was the fact that the experimenters had designed the mouse to fail or "stick" at irregular intervals during the game play. Thus, as subjects competed for the fastest score, their efforts were thwarted (and therefore frustrated) by the failure of the machine. Throughout the experiment the subjects wore sensors on their non-dominant hand and shoulder which recorded their physiological response to our manipulation. It was assumed that the failure of the machine would produce a full-body physiological reaction -- a "frustration response" -- which would lend itself to analysis with the appropriate technology. The data analysis from this experiment is the subject of another Affective Computing research project entitled "Frustration Detection in Human-Computer Interfaces".
A schematic of the frustration experiment design. Two separate systems, the sensing system (run by the Toshiba laptop, top left) and the game system (run by the Power Macintosh, next to the main monitor), are tightly synchronized via a hacked mouse that sends a signal to both systems each time the mouse button is clicked.
Please send email to: email@example.com. This page was last updated October 15 1997.